Dust, as seen through Sanfilippo's microscope

Dust, as seen through Sanfilippo’s microscope.
Microscope images courtesy of Tom Sanfilippo. Other photos by Jon Christian.

It sprawled on the concrete floor of the Artisan’s Asylum: a jumble of beige cabinets bristling with electronics and chilling equipment, a vacuum imaging chamber, and an avalanche of knobs, switches, and indicator lights that evoked an infectious disease thriller. The scanning electron microscope had been neglected for years, falling into disrepair so serious that nobody knew quite which parts were broken.

Tom Sanfilippo intended to fix it. Himself.

Sanfilippo’s most recent day job was as a software engineer for Microsoft, but a few years ago he started taking classes at the Harvard Extension School in nanofabrication and nano analysis. He became captivated by how scanning electron microscopes use particle beams to magnify by hundreds of thousands of times the microcosmos of insects, dust, microorganisms, and other inaccessibly tiny subjects into resplendent, monochromatic landscapes. There is a “Zen to electron microscopy,” he says. Sanfilippo ended up joining Harvard’s Center for Nanoscale Systems as an external user so he could use their imaging equipment.

At Harvard, Sanfilippo met Nicholas Antoniou who was, at the time, responsible for maintaining and training users on the center’s imagining equipment. Antoniou noticed Sanfilippo’s unusual level of interest in the equipment and, last year, he made him an unusual offer: If Sanfilippo would pay the moving costs for an obsolete, nonfunctioning LEO 982 that Harvard no longer wanted, he could have it for free. “They wanted someone to haul it out of here, basically,” Sanfilippo says.

This is not a normal situation, though it’s not unheard of either. Universities and hospitals sometimes need to dispose of large equipment and it makes more sense to offer it to hobbyists than to discard it. In fact, Sudo Room, a hackerspace in Oakland, California, was offered a scanning electron microscope of its own, but had no particular use for it. And just across the bay, San Francisco’s Noisebridge had one, a TV Mini-SEM from International Scientific Instruments, until they discarded it. Scanning electron microscopes are big, unwieldy, and take a lot of effort to restore and run.


Still, any electron microscope is a highly specialized piece of laboratory equipment, and Sanfilippo’s was so old that it ran DOS and sported a 3.5-inch, 1.44MB* floppy drive. It probably retailed for half a million dollars new, and refurbished models still sell for a hundred thousand dollars or more. And, in the offer’s most quixotic twist, nobody seemed to know precisely what was wrong with it — so Sanfilippo didn’t know whether it needed a few new parts or whether every main subsystem was fried, and there was no way to find out without taking it home, opening it up, and getting his hands dirty. Even worse, he knew that if he inadvertently damaged one of the unit’s delicate and specialized components, it could render the elegant machine permanently unfixable.

An electron microscope’s infrastructure requirements alone are intimidating. “It’s not something you can put in a living room,” Antonious says. The LEO 982 uses three kilowatts of power, provided by a 220V supply, and requires sources of compressed air, water, and nitrogen. He investigated the resources at the Artisan’s Asylum, a large makerspace in Somerville, Massachusetts where he serves as chair of the board, and found they could provide much of what he needed. What was less sure was whether he was personally up to the task.

In the end, he decided to go for it, in large part due to the simple novelty of the offer.

“I thought, no one would ever ask me that question again in my life, so I said yes,” he says.

Disassembling and Troubleshooting

First, Sanfilippo had to disconnect the tangle of serial, parallel, SCSI, and coaxial cables that connected the microscope’s cabinets. The machine was located inside a clean room, so he needed to wear a bulky suit while he took it apart. Then there was the sheer complexity of the thing: Dozens of electronic connections and water lines run between the computer terminal, vacuum chamber, and power supply through four corrugated hoses between the units. If he reattached just one incorrectly, it could prevent the device from powering on — or damage it permanently.

In total, Sanfilippo says he spent about three days inside the clean room, photographing and tagging each cable before he disconnected it. Then he hired Harvard’s internal moving service to transport them to the Artisan’s Asylum — a journey, fortuitously, of less than two miles.

Back in Somerville, he surveyed the mound of equipment and service manuals, which were printed in German by LEO owner Carl Zeiss AG.

“Once I got it here, I started discovering problems,” he says. “And then there were more problems that I created when I disassembled it.”


To start, there was a gaping hole in the chamber, which needs to be almost a total vacuum for the electron gun to operate. At Harvard, it had been fitted with an energy-dispersive X-ray spectrometer to perform elemental analysis of samples in the magnification chamber, but before Sanfilippo got to the microscope, someone removed the device, leaving a hole in the vacuum chamber about the diameter of a can of Pepsi. “The poor machine had been picked over quite a bit,” he says.

He decided to build a plate to cover the hole. He took precise measurements, milled it out on the Artisan’s Asylum CNC machine, and fitted it with a rubber seal. It seemed to fit snugly, but to test it, he would need to repair the pumps.

To pull the ultra vacuum the electron gun requires, the LEO 982 uses three pumps. The first, a conventional vacuum, sucks most of the air out of the chamber. Then a turbomolecular pump knocks out individual molecules with a rotor, and finally, a titanium sublimation pump coats the interior of the chamber with a layer of clean titanium that reacts with stragglers as they strike the walls, turning them solid.


Sanfilippo used a USB-to-serial converter to run diagnostics on the turbomolecular pump. The results were grim; he considered rebuilding it, but opted for a reasonably priced refurbished model instead.

He rebuilt and cleaned the conventional pump, though. It was missing a filter, causing it to leak oil through the lines and into the turbomolecular pump, which may have caused that device’s catastrophic damage. He replaced the soiled hoses, installed a new filter, and the pump fired up.

But when he turned on the conventional and turbomolecular pumps, the chamber didn’t pull a vacuum. The new plate he had built was leaking air into the chamber. “I guessed the depth of the channel, and I cut too deep,” he says.

He considered his options. He could go back to the CNC machine and mill a new plate, adjusting the channels to achieve a better seal. Or he could buy a can of Vacseal and spray it around the edges until the imperfect plate held. He opted for the second, and it worked. After applying a thick layer of adhesive, the chamber pulled a vacuum.

Sanfilippo next opened the chiller, which recirculates water at a moderate 68–72°F to cool the high voltage supply and electromagnets, and found the hoses clogged with organic material. “It had filled with algae,” he says. He blasted compressed air through the hoses to clear the chiller, then set about hooking it to the water supply.

But Sanfilippo knew there would be limitations to the facility’s infrastructure. The Artisan’s Asylum is a sprawling, 40,000-square-foot makerspace that houses robots like Project Hexapod and the Autonomous Combat Robots Design Challenge, the bike gang SCUL, and bicycle frame builder Paul Carson; it has comprehensive machining and welding equipment, a uPrint SE Plus, and a Lasersaur 100W laser cutter. But initially, the space where Sanfilippo set up the microscope didn’t even have a water supply. So he cleared acrylic resin out of a clogged sink in a nearby common area (“I became kind of a folk hero for fixing the sink,” he says) and snaked a pair of hoses between the sink and the microscope’s cooling system.

At this point, the chiller and computer should have powered on, but something was wrong. Sanfilippo checked and double-checked the ropes of cabling between the components, flipped through the service manuals, and consulted with electronics experts at the Asylum. He eventually discovered that he had reconnected a cable between the ion sublimation pump’s high voltage supply and the main console backward, which had blown a handful of soldered-on fuses. He found suitable replacement fuses lying around the Artisan’s Asylum, a friend helped him replace them, and the computer powered on.

With the chiller functioning, Sanfilippo could activate the ion sublimation pump. It powered on, and the chamber achieved an ultra vacuum.


The computer started up as well, with only minor hiccups, and booted to MS-DOS 5.22 (“for a hackerspace, running DOS is fine,” Sanfilippo says). He let the electron gun stabilize for a few hours, set the aperture and calibrated the magnetic field, and with a sense of overwhelming relief, found that it was providing an image right away.

“I didn’t know how far I’d get,” he says. “I couldn’t believe it when I got those images.”

Nobody was more surprised than the microscope’s old caretakers at the Center for Nanoscale Systems.

“I suspected that he would need some professional help, and I did give him some contacts,” Antoniou says. “I have decades of experience with these instruments, and I would probably have at least consulted with a service engineer.”

Harald Hass, a representative of Zeiss, which acquired LEO in 2004, was similarly incredulous. “This was for sure a challenge for this person and I think he couldn’t do it without any assistance or support from an experienced engineer,” he says.

You Break It…

That might have been the end of Sanfilippo’s restoration (“I was kind of basking in the success of getting it running,” he says) if he hadn’t left the machine on while he went out of town for his niece’s high school graduation. Remember how the chiller was hooked up to a sink? Over the weekend he was gone, somebody disconnected the hose, and left it disconnected, so the microscope was running without chilled water. Sanfilippo’s best guess is that the unit hit 72 degrees and shut down automatically, but not before overheating electronics caused a hose to burst inside the cabinet.

When he powered it back on, to his horror, water started spraying into the electrical gear and pooling on the floor.

He turned it off, took the cabinet apart, and dried each part with a hair dryer. Realizing the sink wasn’t a long-term solution, he installed a dedicated water line to the chiller. There didn’t seem to be any permanent damage to the electronic equipment, but the episode apparently deformed the tip of the electron gun, which no longer provided a clear image after the overheating incident. Sanfilippo found a steep discount on a replacement electron gun, but there was no documentation on how to replace it. He took the imaging chamber apart carefully, improvising each step, and to his profound relief the installation was successful.


An Unknown Future

On a typical day, the repaired microscope is covered in a thick layer of tools, papers, packaging, and miscellaneous electronics. What happens now that it’s restored? “That’s the next step: What can people do with it?” asks Artisan’s Asylum President Derek Seabury. “Are we going to be doing structural analysis of broken down components, to see if they match up with computer models, or of filaments from 3D printers? It’s going to be a fantastic new tool in the arsenal.”

Sanfilippo plans to teach at least one class on it — with that training, he hopes, others at the Artisan’s Asylum will find uses for it in their own projects — and he’s been talking with the nearby Aeronaut Brewery about using it to enrich a course on beer making. A sobering possibility, Sanfilippo admits, is that he could sell it to help cover the Asylum’s expenses. It’s very hard to tell what a microscope like this would fetch, but others like it have been listed at more than $100,000.

To Sanfilippo, though, the payoff isn’t the finished product; it was the process of painstakingly researching and repairing every key component — and gaining insight into electron microscopy usually reserved for manufacturers and professional technicians — that made the project worthwhile.

“I didn’t know how [an electron microscope] worked,” he says. “I only knew things theoretically. Now I feel like I really know this device.”

*A previous version of this story referred to a “1.44-inch floppy drive,” which is incorrect.